Gridlocking, or sensor registration, is part of a process for combining objects (tracks) from more than one sensor in a consistent manner. This process, known as the correlation and gridlocking process, is intended to provide a single integrated picture (SIP) for the end user. The goal of gridlocking is to determine the difference in position and orientation of the sensors, thereby reducing navigation errors and sensor misalignment errors so that one sensor's track data is accurately transformed into another sensor's coordinate system, e.g. radar-aligned.
One problem involved with gridlocking more than one sensor covering adjacent and overlapping areas or volumes of space is that each sensor typically determines the positions of objects based upon the distance of the object from that sensor and the orientation of that sensor, meaning it is a relative measurement as that coordinate frame may be in motion relative to some other inertial frame. The exact distance and orientation of each sensor may not be available, especially in applications wherein one or more sensors may change positions over a period of time. For example, a ship may comprise a radar and the ship may move forward and roll, causing the position and orientation of the radar to change over time. A possible solution on the radar end is to provide a global positioning system and gyroscope with the radar. The global positioning system provides an absolute position for the radar usually within 0.2 to 100 meters in the ellipsoidal plane of the earth (much less accuracy in altitude) and the gyroscopic system may provide an orientation within some degree of accuracy (from milliradians to degrees depending upon the application). However, providing real-time positions and orientations for more than one radar on the radar end has a relatively significant margin of error for many applications as the probability distributions of these radars errors is difficult to characterize, leading to sub-optimal filter performance in tracking these coordinate frame differences.
The output of a gridlocking or sensor registration process provides positional adjustments (PADs) for tracks to a system (such as a computer with a real-time, updateable geographic map) to model the theater, or the geographical area of interest, which encompasses a greater area than that of an individual radar system. Furthermore, these tracks are aligned with respect to some common coordinate frame with a smaller amount of bias error than would normally be possible without any gridlocking facility. Data is collected by a number of radars for the positions of planes, ships, ground vehicles, troop movements, or other targets. An increase in the data that is collected and processed about positions of the objects at a given moment in time can potentially increase the accuracy of a model or presentation of the positions of the objects in the theater. However, a limitation on computing power and time to describe the positions of the objects in the theater in a coherent way significantly impacts the amount of data that is processed and vice versa.
In a closed-loop correlation and gridlocking system with feedback, computers such as Cray computers, reduce the error in positions of objects within the theater with respect to one another by searching for the minimum of a complicated surface describing the error in registration. Most statistical techniques describe the bias error in a mean-square error sense. There are numerous techniques for minimizing this mean-square error, but the final result in any approach is an optimum location in this parameter space of possible rotations and offsets corresponding to the minimum error. In closed-loop gridlock operation, the bias vector is tracked in a similar manner to that of the state of any given track. Typically, the bias error vector is estimated with the aid of a Kalman filter. This technique requires knowledge of the probability distribution of the measurement noise and process noise as well as characterization of the measurement to state transfer function and state-space matrix, which attempts to describe how the bias error vector evolves with time. Using a predictive-corrective filter structure, the Kalman filter is able to combine bias error vector input from individual track pair combinations and use this information in a manner that factors into account the uncertainty in the input. This is seen as just a real-time implementation of the scheme developed by Gauss at the turn of the 19th century in the pursuit to estimate the position of stellar objects, known as the least-square problem with weighted data. The result is a normalized or average coordinate transformation including a translation and rotation correction that is calculated for transforming tracks of the first radar to the same coordinate system, within a margin of error, as the tracks of the second radar. By normalizing the coordinate transformations of the tracks cross-correlated individually, however, the coordinate transformation includes track data from noisy or erroneous tracks. For example, the Cray computer receives the position and velocity of a plane from the first radar based upon or with respect to a position and azimuth from a global positioning system and compass system for the first radar. The computer system also receives a position and velocity for the object from a second radar wherein the object is determined to be the same plane. By statistical techniques, the Cray computer determines a distance correction and azimuthal or rotational correction to minimize the difference in position and velocity reported for the object by the two tracks. After two or more distance corrections and rotational corrections are calculated, the filter effectively normalizes the distance corrections and rotational corrections. That is, the state of the actual offset and rotational corrections needed must be estimated from the noisy measurements, with the measurements being derived from the set of determined track pairs. However, current systems incorporate error involved with incorrect determinations that the tracks describe the same object. Further, the computing power and time to evaluate each track individually is significant and increases with the square of the number of tracks to compare.
Skilled artisans appreciate that elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale. For example, the dimensions of some of the elements in the figures are exaggerated relative to other elements to help to improve understanding of embodiment(s) of the present invention